CN107608936A - A kind of epicyclic gearbox combined failure feature extracting method - Google Patents

Abstract

The invention discloses a method for extracting complex fault features of a planetary gearbox. Firstly, the vibration signal of the planetary gearbox is measured and stored; secondly, a multi-wavelet symmetric lifting frame is constructed, and control parameters are introduced; then, multi-fractal entropy is constructed as an adaptive matching criterion The evaluation index, through the adaptive construction of multi-wavelet by intelligent optimization algorithm, obtains the multi-wavelet basis function matching with the dynamic signal; then decomposes through redundant multi-wavelet transform; finally, calculates the relative fault characteristic frequency in each frequency band Energy ratio, obtain frequency band relative energy ratio histogram, select fault sensitive frequency band, and then identify and separate compound faults. The invention can overcome the complex transmission path of the planetary gearbox and the influence of working condition noise, and extract and separate the early fault features of the inner ring gear, the planetary gear and the sun gear of the planetary gearbox by means of the self-adaptive multi-wavelet structure and the selection of sensitive characteristic frequency bands.

Description

A Feature Extraction Method for Compound Faults of Planetary Gearbox

technical field

The invention relates to the technical field of planetary gearbox fault detection, in particular to a method for extracting a compound fault feature of a planetary gearbox.

Background technique

Planetary gearboxes are small in size, light in weight, large in transmission ratio, high in efficiency, and strong in carrying capacity. They also have the advantages of multi-force convergence or single-force dispersion. They are widely used in various industries such as new energy vehicles, wind power generators, aerospace, and ships. system. Because planetary gearboxes often operate under harsh environmental conditions such as high speed, heavy load, and strong impact, key parts such as sun gears, planetary gears, and ring gears in planetary gearboxes are prone to wear, fatigue, broken teeth, and cracks. This kind of failure will further induce other failures, resulting in huge economic losses. Therefore, it is of great engineering significance to monitor the operating state of the planetary gearbox and identify its faults in time.

Planetary gearboxes include sun gears, planetary gears and ring gears. During operation, the sun gear meshes with multiple planetary gears at the same time, which is most prone to fatigue and crack damage. In a planetary gearbox, multiple pairs of gears mesh at the same time, and the position of the planetary gears changes continuously, which leads to the constant change of the position and direction of the meshing force of the gear pairs, and the continuous change of the signal transmission path, resulting in a planetary gearbox with a simple structure and a fault mechanism and spectrum structure. Compared with the traditional fixed-axis gearbox, the fault diagnosis problem of complex mechanical system has its own many characteristics and difficulties. At present, the traditional gearbox fault diagnosis theory and technology cannot effectively solve many problems faced by the crack identification of planetary gearboxes.

In view of the characteristics and difficulties of planetary gearbox fault diagnosis, it is necessary to rely on effective fault diagnosis techniques and methods to achieve the purpose of accurately extracting fault features and accurately quantitatively diagnosing faults. Wavelet transform is known as "mathematical microscope" and is a powerful tool for extracting weak features in non-stationary signals. Its physical essence is to search for the components that are most similar or most related to the "basis function" in the signal. However, it has only one basis function, which has inherent shortcomings in fault matching. As a new development of wavelet, multi-wavelet has many excellent properties that single wavelet cannot have at the same time, and has multiple time-frequency characteristics at the same time. The basis function makes the multi-wavelet have significant advantages in the extraction of crack weak features and multiple composite features. However, the fixed multi-wavelet basis function still cannot achieve the optimal match with the weak or multiple compound features of the planetary gear transmission system crack, which limits its ability to extract the crack feature of the planetary gear transmission system.

Contents of the invention

The present invention aims to solve the problem that the traditional gearbox fault diagnosis method cannot effectively solve the problem faced by the crack identification of the planetary gearbox, and provides a compound fault feature extraction method of the planetary gearbox.

In order to solve the above problems, the present invention is achieved through the following technical solutions:

A method for extracting complex fault features of a planetary gearbox, comprising the following steps:

Step 1. Use an acceleration vibration sensor to pick up the vibration signal of the planetary gearbox, and the acceleration sensor is installed on the input shaft end cover of the planetary gearbox to be tested;

Step 2, carrying out adaptive construction of multi-wavelet basis functions to the collected vibration signal;

Step 2.1, select the initial multi-wavelet and other wavelet basis functions and multi-scale function translations used to modify the multi-wavelet, use the symmetry condition and vanishing moment condition to construct the lifting linear equation system, and solve the linear equation system under the underdetermined condition to obtain Lifting coefficient, introduce adjustable free parameters in this process, substitute the lifting coefficient into the lifting coefficient equation and perform Z transformation to realize the symmetrical lifting of multi-wavelet, and complete the construction of multi-wavelet basis function library;

Step 2.2. According to the structural characteristics of the fault characteristics of the planetary gearbox, the normalized multifractal entropy is constructed as the evaluation index in the adaptive optimization process of the multi-wavelet basis function, and the optimal multi-wavelet basis suitable for the fault characteristics is obtained based on the intelligent optimization algorithm function;

Step 3, using the optimal multiwavelet basis function to perform redundant multiwavelet transform on the collected signal to obtain multiple decomposed sub-frequency bands;

Step 4, calculating the relative energy ratio at the fault characteristic frequency of each sub-band, and using the sub-band with relatively high relative energy as the sensitive sub-band where the composite fault feature is located;

Step 5. Perform Hilbert envelope demodulation processing on the sensitive sub-bands obtained from the decomposition one by one, extract the relevant features of the compound fault of the planetary gearbox, and carry out identification and diagnosis.

In the above step 2.1, the free parameters are determined by solving the solution of the lifting coefficient equation under the underdetermined condition.

The specific process of the above step 2.2 is as follows:

First, select the mother wavelet with appropriate vanishing moment, determine the partition function, and describe the scale behavior of the partition function through the quality factor, and then perform Legendre transformation on the quality factor to obtain the multifractal spectrum of the signal;

Secondly, the information entropy calculation is introduced into the multifractal spectrum, the multifractal entropy is constructed, and the normalized multifractal entropy is obtained by normalization processing;

Finally, based on the intelligent optimization algorithm, the minimum sum of normalized multifractal entropy of the detail signals corresponding to the two wavelet basis functions after multiwavelet decomposition is optimized as the optimization objective function, and the adaptive multiwavelet basis functions are obtained.

In the above step 2.2, the intelligent optimization algorithm is a genetic algorithm.

In the above step 2.2, the normalized multifractal entropy H _n is:

In the formula, f(α _i ) is the multifractal spectrum of the signal.

In the above step 3, the number of redundant multi-wavelet transform layers for multi-wavelet basis functions is 3 layers.

In the above step 4, the relative energy ratio r at the fault characteristic frequency of each sub-band is:

where f _c ∈ (f _c - Δ, f _c + Δ)

In the formula, A is the amplitude of the square envelope spectrum, f _c is the characteristic frequency, Δ is the selected frequency interval, and f=0~f' is the sub-band range.

Aiming at the fact that the composite fault of planetary gearbox contains two or more different fault features, and the fault characteristic waveforms of different fault types are different, the present invention utilizes multi-wavelets to have multiple wavelet basis functions, which has natural advantages in the extraction of composite fault features. At the same time, to avoid the fact that the fixed multi-wavelet basis function cannot achieve the best match with the fault characteristic waveform, which is not conducive to the optimal extraction of fault features, an adaptive multi-wavelet construction method based on multifractal entropy is proposed, which has the following significant advantages:

1. The present invention overcomes the difficult problem of compound fault diagnosis of planetary gearbox, utilizes the characteristics of multiple wavelet basis functions in multi-wavelet, and multiple wavelet basis functions are matched with multiple fault features, so that the compound fault feature extraction and separation of planetary gearbox become possible;

2. The present invention uses fractal geometry to capture the irregular singularity signal with geometric structure characteristics induced by local abnormalities of mechanical equipment, constructs normalized multifractal entropy as an optimization index, and utilizes normalized multifractal entropy after multi-wavelet decomposition The minimum sum principle guides and optimizes the best multi-wavelet basis function suitable for composite fault features, and obtains excellent composite fault feature extraction and separation capabilities;

3. With the help of the calculation of the relative energy ratio at the fault characteristic frequency in each frequency band after multi-wavelet decomposition, the relative energy ratio histogram of the frequency band is obtained, which intuitively shows the fault characteristic energy of each component of the planetary gearbox, which is beneficial to sensitive The choice of eigenfrequency;

4. The present invention can be used for planetary gearbox fault diagnosis based on vibration monitoring, and can extract early complex fault features of sun gear, ring gear and planetary gear, avoid sudden accidents in the transmission system, and reduce economic losses.

Description of drawings

Fig. 1 is a kind of flow chart of the diagnostic method of compound fault of planetary gearbox;

Figure 2 is a time-domain waveform diagram of the vibration signal of the planetary gearbox;

Fig. 3 is the spectrum diagram of the original signal of the planetary gearbox;

Fig. 4 is the envelope spectrum diagram of the original signal of the planetary gearbox;

Figure 5 is the adaptive multi-wavelet basis function of the planetary gearbox vibration signal; where (a) is the multi-wavelet function ψ ₁ , (b) is the multi-wavelet function ψ ₂ ;

Fig. 6 is a histogram of multi-wavelet packet decomposition energy ratio of planetary gearbox;

Figure 7 is the envelope spectrum diagram of the sensitive frequency band branch; where (a) is the envelope spectrum of the 7th branch, and (b) is the envelope spectrum of the 16th branch.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in combination with specific examples and with reference to the accompanying drawings.

A method for extracting complex fault features of a planetary gearbox, as shown in Figure 1, specifically includes the following steps:

The first step: signal acquisition.

A vibration acceleration sensor is used to pick up the vibration signal of the planetary gearbox, and the acceleration is installed on the input shaft end cover of the planetary gearbox to be tested.

The second step: adaptive construction of multi-wavelet basis functions for the collected vibration signals.

First of all, the multi-wavelet basis function is constructed using the multi-wavelet symmetric lifting framework, which ensures that the filter has a linear phase or a generalized linear phase through symmetry, which is beneficial to boundary processing and avoids phase distortion when decomposing and reconstructing the signal; In this process, adjustable free parameters are introduced; secondly, according to the structural characteristics of the fault characteristics of the planetary gearbox, the normalized multifractal entropy is constructed as the evaluation index in the process of multi-wavelet basis function adaptive optimization, which is obtained and compounded based on the intelligent optimization algorithm. Adaptive Multiwavelet Basis Functions to Fault Features.

(1) Using the lifting framework to construct multi-wavelet basis functions.

First, select the initial multiwavelet ω ₀ (x), where ω ₀ (x)=ψ ₁ or ψ ₂ ; then select other basis functions ω ₁ (x),...,ω _k ( The translation amount k of x). Finally, a new multi-wavelet can be constructed through the "lifting coefficient equation" The lift coefficient equation is:

Compared with single wavelet, multi-wavelet has more advantages in lifting construction. For example, in single wavelet lifting, the scaling function can only be used to modify the original wavelet function, that is, ω _i in the above formula can only be In multi-wavelet lifting, the correction of a certain multi-wavelet function includes not only two multi-scale functions, but also another corresponding multi-wavelet function, that is, for ψ ₁ , for Obviously, the basic functions used to construct new multi-wavelet functions are more than single wavelets, which brings more freedom and flexibility to the construction of multi-wavelets to meet more and more specific requirements.

Assuming that the vanishing moment of the multi-wavelet is raised from p to p', the following lifting linear equations can be obtained by integrating both sides of the "lifting coefficient equation".

The solution {c _i } of the equation system is the coefficient of the multi-wavelet lifting function. Performing Z-transform on Equation (1) can obtain the multi-wavelet lifting frame.

(2) Perform symmetric lifting of multi-wavelets.

In order to ensure the symmetry of the lifting process, the "symmetry selection" method is used to select the translation amount of other functions used to modify the multi-wavelet. Assuming an initial multiscale function It is symmetric or antisymmetric to the multiwavelet functions ψ ₁ and ψ ₂ , and the symmetry centers are respectively Then the symmetric lifting method is expressed as follows, taking the symmetric lifting of ψ ₁ as an example, the translation of the lifting function Must meet:

In the formula: i=1,2; j=1,2,...; m=1,2.

make Respectively represent the symmetry properties of the initial multiscale function and the initial multiwavelet function, where 1 represents symmetry and -1 represents anti-symmetry. Will and M(ψ _i ,k,n)= _∫ψi (x+k)x ⁿ dx are substituted into formula (2), and the first matrix on the left side of the equal sign represents M _B , where M _B =MB, M and B respectively

Let B be a symmetric matrix

The coefficient vector in formula (2) is expressed as And the right side of the equation is expressed as M _ψ =[M(ψ _i ,0,p),M(ψ _i ,0,p+1),…M(ψ _i ,0,p'-1)] ^T , then the formula (2) becomes the following formula

M _B C = M _ψ (6)

The solution C of the equation is the coefficient used to lift ψ ₁ , the situation of ψ ₂ is similar, the only difference is that the function of lifting ψ ₂ is and Substitute the lifting coefficient into formula (2), and perform Z transformation to obtain the lifting matrices T and S. Therefore, the multi-wavelet symmetric lifting construction can be realized with the help of lifting matrices T and S, as follows:

In the formula: H _new is the low-pass filter symbol after multi-wavelet symmetric lifting construction; G _new is the high-pass filter symbol after multi-wavelet symmetric lifting construction.

(3) Obtaining free parameters in adaptive construction.

There are three situations in the solution of lifting coefficient equation (6): (a) the equation is overdetermined and has no solution; (b) the equation is positive definite and has only one solution; (c) the equation is underdetermined and has multiple solutions. The flexibility and freedom of using the lifting method to construct multi-wavelets come from the free parameters in the lifting process, which requires the solution of the equation to be the third type. Reducing the matrix M _B in formula (6), the original linear equation system can be reduced to a non-redundant linear equation system, the number of equations is Rank(M _B ), and Rank refers to the rank of the matrix, so the linear equation There may be N _f =(p'-p)-Rank(M _B ) free parameters in the equation system. Since the essence of self-adaptation is the process of optimizing the included free parameters in accordance with the specified objectives through some optimization method, it can be seen that these free parameters are the key to realize multi-wavelet self-adaptation.

(4) Optimize the target structure.

The signals induced by mechanical equipment due to local abnormalities often have singularity, which manifests as irregular transient structures such as mutations and sharp points. A multifractal is a set of singular measures defined in the fractal scale-free interval with multiple scale indices. It describes the distribution of fractal measures on the support set, and spectral functions can be used to describe different levels of fractals. Characteristics. In order to accurately identify the fault type and quantify the damage degree, the information entropy is combined with the multifractal spectral function to form multifractal entropy. The multifractal entropy is introduced into the adaptive multiwavelet construction as the objective function of the optimization process, and the calculation of the multifractal entropy is as follows:

Suppose f(x) is a function of finite energy, that is, f(x)∈L ² (R), then the wavelet transform definition of this function is shown in the following formula:

In the formula: ψ _a,b (x) is the mother wavelet.

When the scale a=a ₀ , if any point b in a certain field of b=b ₀ has:

|W _f (a ₀ ,b;ψ)|≤|W _f (a ₀ ,b ₀ ;ψ)| (9)

Then (a ₀ , b ₀ ) is called the local maximum modulus point, and the line connecting all the maximal modulus points in the scale space is called the maximal modulus line. The size and position of local maxima in wavelet transform contain rich signal information. An important property of the local maximum modulus of wavelet is that when the Hausdorff exponent α of the analyzed signal f(x) at a certain point x ₀ is smaller than the vanishing moment of the wavelet, there is at least one local maximal modulus pointing to this point, and The following scaling behavior exists for the wavelet transform coefficients along the maximal modulus:

The fractal feature of the signal is mainly manifested as the hierarchical distribution of singularity, and the hierarchical structure of singularity can be highlighted through the wavelet local maximum modulus.

The key to calculating the multifractal spectrum with the wavelet maximal modulus is the representation of the partition function. Let the set of all wavelet maximal modulus lines below the scale a be L(a), then define the partition function on this set as:

The supremum of the maximum value of the wavelet modulus is selected by the above formula, which overcomes the instability problem of Z(q, a) caused by the small value of the wavelet transform modulus near the point on the maximum value line. At the same time, it also avoids the problem of rapid data increase caused by the maximum value generated by the rapid oscillation of Z(q,a). But when the scaling function a→0, the scaling behavior of the partition function can be expressed by the quality factor τ(q)

Z(q,a)=a ^τ(q) (12)

By performing Legendre transformation on the quality factor τ(q):

Thus the multifractal spectrum f(α) is obtained as:

Let f(α _i ), 1≤i≤k be the multifractal spectrum of the signal, and the size of f(α _i ) reflects the proportion of the fractal dimension under the corresponding singularity index in the total fractal dimension. According to the measurement formula of entropy, there is multifractal entropy:

In the formula:

The normalized multifractal entropy _Hn is defined as:

It can be seen from the above formula that H _n ∈ [0,1]. When a device fails, the signal will have different degrees of singularity. The sharper the fault impact, the smaller the singular value; while entropy is a measure of certainty and complexity, the signal with singularity has stronger periodicity and self-similarity. The better it is, the more certain the state of the equipment is. At this time, the normalized multifractal entropy H _{n is} smaller and closer to 0.

(5) Adaptive structure.

In view of the strong robustness of the genetic algorithm and the characteristics of global and parallel search, it does not need the mathematical expression between the objective function and the variables. The present invention adopts the genetic algorithm as the optimization algorithm, and the normalized multifractal entropy constructed is used as the optimization evaluation index to obtain the normalized multifractal entropy of the detail signals corresponding to the two wavelet basis functions after multiwavelet decomposition; The optimal problem of wavelet basis function is transformed into the optimization problem of the minimum sum of normalized multifractal entropy corresponding to the detail signal after decomposing two multi-wavelet basis functions of the objective function; obtain the optimal multi-wavelet basis function suitable for composite fault characteristics .

Step 3: Obtain sensitive characteristic frequency bands.

Multiple decomposed sub-bands are obtained through redundant multi-wavelet transform. The relative energy ratio at the fault characteristic frequency of each sub-frequency band is calculated, and the frequency band where the composite fault is located is intuitively obtained through the relative energy ratio histogram.

Using the optimal multi-wavelet basis function, 2·2 ³ =16 branches are generated after 3-layer redundant multi-wavelet packet decomposition; at the same time, the multi-wavelet reconstruction process is discarded to obtain multi-branch information expression, and the multi-branch information expression is obtained by formula (17). After wavelet decomposition, the relative energy ratio at the fault characteristic frequency in each frequency band is obtained, and the relative energy ratio histogram of the frequency band is obtained. Each multi-wavelet basis function is decomposed into 8 sub-bands, a total of 16 sub-bands; the distribution of faults is reflected from the figure Find the optimal sensitive frequency band in the situation for analysis.

where f _c ∈ (f _c -Δ,f _c +Δ) (17)

In the formula: f _c is the characteristic frequency/Hz; A is the amplitude of the square envelope spectrum; Δ is the selected frequency interval, and f=0~f' is the frequency band range.

Step 4: Composite fault feature extraction:

The Hilbert envelope demodulation is performed on the decomposed sensitive sub-bands one by one, and the relevant features of the compound fault of the planetary gearbox are extracted and identified and diagnosed.

Below by a specific example, the performance of the present invention is described in further detail:

(1) The acceleration vibration sensor is used to pick up the vibration signal of the planetary gearbox, and the acceleration sensor is installed on the input shaft end cover of the planetary gearbox to be tested.

The parameters of the test planetary gearbox are shown in Table 1.

Table 1 Planetary gearbox parameters

The calculation formula of the fault characteristic frequency of each transmission element of the planetary gearbox is as follows:

In the formula: f _in is the rotational frequency of the input shaft/Hz; f _s is the sun gear; f _p is the fault characteristic frequency of the planetary gear/Hz; f _r is the fault characteristic frequency of the inner ring gear/Hz; Z _s is the sun gear The number of teeth; Z _p is the number of teeth of the planetary gear; Z _r is the number of teeth of the inner ring gear.

During the test, the input speed of the planetary gearbox is 1200r/min, and a vibration acceleration sensor is arranged at the input end of the planetary gearbox to obtain vibration signals, and the sampling frequency is 12.8KHz. Calculated by formulas (18)-(20), the characteristic frequency of the first stage of the planetary gearbox is shown in Table 2.

Table 2 The characteristic frequency of the first stage transmission of the planetary gearbox when the speed is 1200r/min

The time-domain waveform of the vibration signal collected by the vibration acceleration sensor is shown in Figure 2, and the data length is 32768. The spectrum obtained by using the traditional FFT method and the envelope spectrum obtained by Hilbert transform are shown in Fig. 3 and Fig. 4 .

It can be seen from Figure 2 that the time-domain signal is basically submerged by noise, and it is difficult to judge the fault characteristics. From its spectrum diagram 3, it can be seen that the energy is mainly concentrated on the meshing frequency and its multiplier of the first-stage planetary gearbox, and zooming in on the vicinity of the meshing frequency, it can be seen that there are sidebands separated by the characteristic frequency of the inner ring gear fault . From the enlarged figure 4 of the low frequency band of the envelope spectrum, it can be seen that the fault characteristic frequency of the inner ring gear is still obvious. Therefore, it is inferred that the planetary gearbox may have an internal ring gear fault, but other fault features of the planetary gearbox in the original signal spectrum and envelope spectrum are overwhelmed by noise.

(2) Select Hermite as the initial multi-wavelet, raise the multi-wavelet vanishing moment to the fifth order, use the symmetrical lifting frame to carry out multi-wavelet lifting, obtain free parameters, and use the genetic algorithm to decompose the normalized multiple analysis entropy and the minimum sum is Optimizing the objective and obtaining the optimal multi-wavelet basis functions are shown in Figure 5, Figure 5(a) is the multi-wavelet function ψ ₁ , and Figure 5(b) is the multi-wavelet function ψ ₂ .

(3) Due to the diversity and complexity of the compound fault features of planetary gearboxes, its fault features may be distributed in different time-frequency positions. In order to fully and completely reflect the analysis results of compound faults, firstly, the optimal multi-wavelet basis function is used After three layers of redundant multi-wavelet packet decomposition, 2·2 ³ =16 branches are generated; secondly, the multi-wavelet reconstruction process is discarded to obtain multi-branch information expression, and the fault characteristic frequency in each frequency band after multi-wavelet decomposition is calculated The relative energy ratio at the position, obtain the frequency band relative energy ratio histogram, as shown in Figure 6, the first 8 branches in the figure correspond to the frequency bands of the first multi-wavelet basis function decomposition, and the last 8 branches correspond to the second multi-wavelet basis function decomposition Find the optimal sensitive frequency band from the distribution of faults reflected in the figure for analysis.

It can be seen from Figure 6 that the relative energy of planetary gear failure and sun gear failure is concentrated in the 7th branch, while the relative energy of the ring gear failure is the highest in the 16th branch. Select the 7th branch and the 16th branch as sensitive frequency bands, and perform Hilbert envelope spectrum on them, as shown in Figure 7. It can be seen from Figure 7(a) that the fault characteristic frequency of the planetary gear is 5.83Hz and the sun gear fault characteristic frequency is 52.5Hz, and the fault characteristic frequency of the inner ring gear is 7.5Hz. In Fig. 7(b), it is obvious that the component of the fault characteristic frequency of the inner ring gear at 7.5 Hz is very prominent. From Figure 7, it can be judged that the inner ring gear has a damage fault, while the planetary gear and the sun gear have a relatively slight damage fault.

Unpacking inspection revealed a 7×0.5mm ² scratch on the inner ring gear, a 3×0.5mm ² scratch and a 2×1mm ² scratch on the planetary gear, and a 2×1mm 2 scratch on the sun gear. Some pitting failures. Thus, the effectiveness of this method in feature extraction of compound faults of planetary gearboxes is verified.

The invention proposes a compound fault feature extraction method of a planetary gearbox. Firstly, the vibration signal of the planetary gearbox is measured and stored; secondly, the multi-wavelet symmetric lifting frame is constructed, and control parameters are introduced; then, the normalized multifractal entropy is constructed as the evaluation index of the adaptive matching criterion, and the multi-wavelet Adaptive construction to obtain the multi-wavelet basis function matching the dynamic signal; then decompose through redundant multi-wavelet transform; finally, calculate the relative energy ratio at the fault characteristic frequency in each frequency band, and obtain the histogram of the relative energy ratio of the frequency band, Select fault-sensitive frequency bands to identify and separate compound faults.

The present invention uses the adaptive redundant multi-wavelet transform to efficiently and reliably decompose multiple sub-band signals through simple vibration signal measurement, and then obtains the sensitive frequency band of the fault feature through the relative energy ratio at the fault characteristic frequency of each sub-band. The Hilbert transform extracts the relevant features of the planetary gearbox and identifies the diagnosis. The invention can overcome the complex transmission path of the planetary gearbox and the influence of working condition noise, and extract and separate the early fault features of the inner ring gear, the planetary gear and the sun gear of the planetary gearbox by means of the self-adaptive multi-wavelet structure and the selection of sensitive characteristic frequency bands.

It should be noted that although the above-mentioned embodiments of the present invention are illustrative, they are not intended to limit the present invention, so the present invention is not limited to the above specific implementation manners. Without departing from the principles of the present invention, all other implementations obtained by those skilled in the art under the inspiration of the present invention are deemed to be within the protection of the present invention.

Claims (7)

1. a kind of epicyclic gearbox combined failure feature extracting method, its feature are being that it specifically comprises the following steps：

Step 1, epicyclic gearbox vibration signal is picked up using acceleration vibrating sensor, the acceleration transducer is arranged on to be measured Epicyclic gearbox input shaft end covers；

Step 2, the self-adaptive construction that multi-wavelet bases function is carried out to the vibration signal collected；

Step 2.1, select initial m ultiwavelet and for correct m ultiwavelet other wavelet basis functions and multi-scaling Functions it is flat Shifting amount, using symmetric condition and vanish-moment construction lifting system of linear equations, solve the system of linear equations owed under fixed condition and obtain Lifting Coefficients are obtained, introduce regulatable free parameter in the process, Lifting Coefficients are substituted into Lifting Coefficients equation and carry out Z changes The symmetrical lifting for realizing m ultiwavelet is changed, completes the construction of multi-wavelet bases function library；

Step 2.2, the design feature according to epicyclic gearbox fault signature, construction normalization multi-fractal entropy is as multi-wavelet bases Evaluation index during function adaptive optimization, obtained and adaptable optimal how small of fault signature based on intelligent optimization algorithm Ripple basic function；

Step 3, the signal gathered using optimal multi-wavelet bases function pair carry out redundancy multi-wavelet transformation, after obtaining multiple decomposition Sub-band；

Step 4, relative energy ratio at the fault characteristic frequency of each sub-band is calculated, by the higher sub-band of relative energy As the sensitive sub-band where combined failure feature；

Step 5, Hilbert envelope demodulation processing is carried out to the sensitive sub-band that decomposition obtains one by one, extract epicyclic gearbox Combined failure correlated characteristic, and carry out identifying and diagnosing.

2. a kind of epicyclic gearbox combined failure feature extracting method according to claim 1, it is characterized in that, step 2.1 In, free parameter is determined by solving the Lifting Coefficients non trivial solution owed under fixed condition.

3. a kind of epicyclic gearbox combined failure feature extracting method according to claim 1, it is characterized in that, step 2.2 Detailed process it is as follows：

First, morther wavelet of the selection with suitable vanishing moment, determines partition function, and describe partition function by quality factor Scaling behavior, and then Legendre is carried out to quality factor and converts to obtain the multifractal spectra of signal；

Secondly, comentropy is calculated and be introduced into multifractal spectra, construct multi-fractal entropy, and be normalized to obtain normalizing Change multi-fractal entropy；

Finally, based on intelligent optimization algorithm, the normalization of detail signal corresponding to latter two wavelet basis function is decomposed with m ultiwavelet The minimum optimization object function of multi-fractal entropy sum optimizes, and obtains adaptive multi-wavelet bases function.

4. a kind of epicyclic gearbox combined failure feature extracting method according to claim 1 or 3, it is characterized in that, step In 2.2, intelligent optimization algorithm is genetic algorithm.

5. a kind of epicyclic gearbox combined failure feature extracting method according to claim 1 or 3, it is characterized in that, step In 2.2, multi-fractal entropy H is normalized_nFor：

<mrow> <msub> <mi>H</mi> <mi>n</mi> </msub> <mo>=</mo> <mo>-</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>k</mi> </munderover> <mfrac> <mrow> <mi>f</mi> <mrow> <mo>(</mo> <msub> <mi>&amp;alpha;</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>k</mi> </munderover> <mi>f</mi> <mrow> <mo>(</mo> <msub> <mi>&amp;alpha;</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> </mfrac> <msub> <mi>log</mi> <mi>k</mi> </msub> <mrow> <mo>(</mo> <mfrac> <mrow> <mi>f</mi> <mrow> <mo>(</mo> <msub> <mi>&amp;alpha;</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>k</mi> </munderover> <mi>f</mi> <mrow> <mo>(</mo> <msub> <mi>&amp;alpha;</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>)</mo> </mrow> </mrow>

In formula, f (α_i) be signal multifractal spectra.

6. a kind of epicyclic gearbox combined failure feature extracting method according to claim 1, it is characterized in that, in step 3, The number of plies that redundancy multi-wavelet transformation is carried out to multi-wavelet bases function is 3 layers.

7. a kind of epicyclic gearbox combined failure feature extracting method according to claim 1, it is characterized in that, in step 4, Each the relative energy at the fault characteristic frequency of sub-band is than r：

Wherein f_c∈(f_c- Δ, f_c+Δ)

In formula, A be squared envelope spectrum amplitude, f_cFrequency is characterized, Δ is selected frequency separation, and f=0~f ' is son frequency Band scope.